Generation and Characterization of the Eμ-Irf8 mouse model

Abstract

In mature B-cell malignancies, chromosomal translocations often juxtapose an oncogenic locus to the regulatory regions of the immunoglobulin genes. These genomic rearrangements can associate with specific clinical/pathological sub-entities and inform diagnosis and treatment decisions. Recently, we characterized the t(14;16)(q32;q24) in diffuse large B-cell lymphoma (DLBCL), and showed that it targets the transcription factor IRF8, which is also somatically mutated in ~10% of DLBCLs. IRF8 regulates innate and adaptive immune responses mediated by myeloid/monocytic and lymphoid cells. While the role of IRF8 in human myeloid/dendritic-cell disorders is well established, less is known of its contribution to the pathogenesis of mature B-cell malignancies. To address this knowledge gap, we generated the Eμ-Irf8 mouse model, which mimics the IRF8 deregulation associated with t(14;16) of DLBCL. Eμ-Irf8 mice develop normally and display peripheral blood cell parameters within normal range. However, Eμ-Irf8 mice accumulate pre-pro-B-cells and transitional B-cells in the bone marrow and spleen, respectively, confirming that this model amplifies the physiologic role of Irf8 in B-cell development. Notably, in Eμ-Irf8 mice, the lymphomagenic Irf8 targets Aicda and Bcl6 are overexpressed in mature B-cells. Yet, the incidence of B-cell lymphomas is not increased in the Eμ-Irf8 model, even though their estimated survival probability is significantly lower than that of WT controls. Together, these observations suggest that the penetrance on the Irf8-driven phenotype may be incomplete and that introduction of second genetic hit, a common strategy in mouse models of lymphoma, may be necessary to uncover the pro-lymphoma phenotype of the Eμ-Irf8 mice.

Introduction

Chromosomal translocations are an integral part of cancer pathogenesis. In hematological malignancies, they play an outsized role and, as exemplified by the t(9;22) /BCR-ABL fusion, and t(15;17) /PML-RARa fusion, these abnormalities can inform the disease biology, diagnosis and the development of rational therapies[1]. In mature B -cell malignancies, chromosomal translocation often lead to the juxtaposition of a “target” gene locus to the regulatory regions of the immunoglobulin heavy (IGH) or light chain loci[2]. These genomic rearrangements result in deregulated expression of a multiplicity of genes, including the transcriptional factors MYC, BCL6, IRF4 and PAX5, the cell cycle regulator Cyclin D1 (CCND1), and the anti-apoptotic product BCL2. In the context of mature B-cell tumors, these are gain of function changes that can associate with specific clinical/pathological sub-entities, e.g., CCND1 and mantle cell lymphoma (MCL), IRF4 and multiple myeloma (MM) or pediatric B cell lymphoma, and “double-hit” (MYC and BCL2) or “triple-hit” (MYC, BCL2 and BCL6) with high-grade B-cell lymphomas that hold especially ominous prognosis[2, 3]. Further, many of the genes targeted by chromosomal translocation are also often somatically mutated, reflecting the selective pressure within the tumor for their deregulation as well as their structural susceptibility to the corruption of physiologic processes that are germane to B cells, including somatic hypermutation (SHM) and class switch recombination (CSR), which can induce double strand DNA breaks outside the Ig locus[4].

Recently, using an innovative next generation-based capture-sequencing strategy, we discovered and characterized the genomic structure of the t(14;16) (q32;q24) in DLBCL and showed that it bring the regulatory elements of IGH (enhancer mu – Eμ or Switch gamma – Sγ) to the IRF8 gene locus[5]. In agreement with the concept that lymphomagenic genes are often targeted by translocations and mutations, IRF8 was also found to be somatically mutated in ~10% of DLBCL biopsies[6–8]. In these instances, the large majority of variants are missense, with clusters in the N-terminal DNA binding and C-terminal IRF association domain.

IRF8 is a member of the interferon family of transcription factors. It regulates innate and adaptive immune responses mediated by myeloid and lymphoid cells[9, 10]. Binding to distinct cofactors ultimately define if the outcome of IRF8 engagement is activating or repressing[9, 10]. In the myeloid lineage, IRF8 controls the fate of progenitor cells by directing monocytic over neutrophilic differentiation. Demonstrating the relevance of the transcription programs regulated by IRF8 in the myeloid lineage, engineered deletion or spontaneous loss-of-function mutation in this gene causes an CML-like syndrome in mice and a rare syndrome of dendritic-cell (DC) immunodeficiency in humans[11–13]. Although the physiologic and pathogenetic roles of IRF8 in myeloid cells and DCs are well defined, less is known about its contribution to B cell biology. Still, recent work has implicated IRF8 in B cell differentiation and in the regulation of mature B cells localization to specific sub-compartments of secondary lymphoid organs[14]. IRF8 also appears to play an important role during the germinal center (GC) reaction for it is highly expressed in reactive lymphoid tissues[15, 16], and directly activates the transcription of critical elements involved in this process, including the induction of BCL6 and AICDA[17], and suppression of PRDM1[18]. These data are particularly important in the context of DLBCL biology because these lymphomas arise from cells that have been exposed to the GC reaction, and they often display deregulation of IRF8 target genes[3, 19]. Lastly, in additional support to the putative role of IRF8 as an oncogene in mature B cell malignancies (above and beyond our recent discovery of IGH-IRF8 fusions and of somatic mutations), a variant in IRF8’s 3’ UTR has been linked to chronic lymphocytic leukemia (CLL) risk[20], and, importantly, a recent genomewide unbiased screen showed a highly skewed loading of the bromodomain 4 protein (BRD4) at the IRF8 enhancer (super-enhancer), that increases IRF8’s expression and activity in DLBCL[21].

Despite these advances, it remains to be defined whether deregulated IRF8 expression in B cells can per se promote lymphoma in vivo. Thus, to start to address the role of IRF8 in B-cell lymphoma biology we generated and characterized a transgenic mouse model, henceforth named Eμ-Irf8, in which B-cell specific Irf8 expression is driven by the murine Eμ enhancer, therefore mimicking the human t(14;16) (q32;q24) that we and others have reported[5, 22].

Materials and Methods

Generation of the Eμ-Irf8 mouse model

Full length murine Irf8 was PCR-amplified from the cDNA clone MGC:6194 (IMAGE:3487214), cloned into pEμ/pBSVE6BK (a gift from L. Godley, U. Chicago, IL) using HpaI (into EcoRV) and Sal, and sequence verified. The pEμ/pBSVE6BK-Irf8 plasmid construct contains the mouse Eμ enhancer and VH186.2 promoter sequences upstream to the murine Irf8 cDNA, and the human β-globin splice-sites/polyA signal downstream to it. The pEμ/pBSVE6BK-Irf8 plasmid was next digested with BssHII, the transgene-containing fragment gel-purified and injected into one-cell C57Bl/6 embryos (pronuclear injection). In total, 240 embryos (C57BL/6J background) were injected (MD Anderson Genetically Engineered Mouse Facility - GEMF) and 15 pups born from two independent litters and screened for the transgene integration.

Eμ-Irf8 husbandry, copy number quantification and genotyping

Founder Eμ-Irf8 females were crossed with C57BL/6J males. The Eμ-Irf8 male progeny from this first cross were then bred with C57BL/6J females. Eμ-Irf8 males from this second backcross were bred to C57BL/6J females, and Eμ-Irf8 males and females from this additional backcross were bred to generate Eμ-Irf8 homozygotes. For genotyping, DNA was isolated from toe or tail clips from 2–3 weeks old mice, and a duplex PCR performed with Eμ-Irf8-directed oligonucleotides anchored at the VH promoter and β-globin splice-sites, and oligonucleotides mapping to the unrelated miR-155 locus serving as DNA integrity (WT) controls, as we reported[23]. For copy number quantification, the transgene construct mass was calculated, fixed amounts spiked in tail genomic DNA, and a standard curve created by PCR. Densitometric comparison of PCR amplicon intensity from standards to that of founder mice provided an estimation of the number of copies of the transgene inserted in each genome. All animal procedures were approved by the Animal Care and Use Committee of the UTHSCSA. Oligonucleotide sequences are in Supplemental Table 1.

Complete blood counts

These assays were performed in two cohorts of mice: a) 8 mice from the founders litter, examined at 3 and 6 months old and, b) 17 mice (WT, Eμ-Irf8 hemizygous or homozygous) with an average age of 15 months. Blood (200μl) was drawn from the retroorbital vein (1st cohort) or facial vein (2nd cohort) and examined in an automated VetScan HM5 Hematology Analyzer, at the Department of Laboratory Animal Resources of the UTHSCSA.

RNA isolation, cDNA synthesis and q-RT-PCR

RNA was isolated from spleen or purified splenic B cells (isolated with the EasystepTM Mouse B cell kit - Stem Cell Technology, cat# 19854) using Trizol (Invitrogen), as we described[24]. One microgram of RNA was used for cDNA synthesis using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems), as reported[25]. Gene expression was measured on the QuantStudio 5 real-time PCR system (ThermoFisher Scientific) with Irf8 specific primers and the Tbp gene as control. The q -RT-PCRs were performed in triplicate and included a no-reverse transcriptase reaction and a no-template control (water) to test for genomic DNA amplification or contamination. Relative gene expression was calculated using the 2^–ΔΔCt method, as showed before[26].

Protein isolation and western blots

Whole-cell lysates were obtained from spleen or purified splenic B cells and bone marrow using the NP40 lysis buffer (1% NP40, 50mM Tris pH8.0, 150mM NaCl, 10% glycerol, 1mM EDTA, with protease/phosphatase inhibitors)[27]. P roteins were separated by SDS/polyacrylamide gel electrophoresis (SDS-PAGE), transferred to polyvinylidene difluoride (PVDF) membrane in 1X Tris-glycine buffer (3.03g/L Tris-base, 14.4g/L glycine, 20% methanol), as reported earlier[28]. For detection of relevant proteins, the following antibodies were used: IRF8 (Santa Cruz Biotechnology, cat# sc-365042, 1:1000), AICDA (Cell Signaling Technology, cat# 4975, 1:1000), BCL6 (Santa Cruz Biotechnology, cat# sc-858 or Cell Signaling Technology, cat# 4242, 1:1000), β-actin (Sigma-Aldrich, cat#A2228, 1:20000). Goat anti-rabbit or mouse IgG-HRP conjugate were used as secondary antibodies (Bio-Rad Laboratories, cat#1706515 and cat#1706516, 1:5000 or 1:10000, respectively). The proteins were visualized using the SuperSignal^® West Pico PLUS Chemiluminescent Substrate (Thermo Scientific, cat# 34580), and manual film development or digital images (FluorChemR system, ProteinSimple).

FACS analysis

Mice were sacrificed and spleens and femurs harvested for isolation of splenic and bone marrow cells, respectively[29]. For both tissues, r ed cells were depleted using the Lysing Buffer Hybri-MaxTM (Sigma-Aldrich, cat# R7757). For the analysis of the spleen cells a 4-color panel was used: anti-CD23-PE/CY7 (Biolegend, cat# 101614), anti-CD21/CD35-PE (Biolegend, cat# 123409), anti-B220-APC (Invitrogen, cat# 17–0452-82) and anti-IgM-FITC (BD Biosciences, cat# 553437). For the subpopulations quantification in the spleen, cells within a B220+/IgM+ gate, were characterized for expression of CD21 and CD23, and defined as: transitional B cells (CD21-, CD23-), follicular B cells (CD21 low, CD23+), and marginal zone B cells (CD21 high, CD23-/low). In the bone marrow, a 5-color panel was used: anti-B220-APC, anti-IgM-FITC, anti-C43-PE/CY7 (Biolegend, cat# 143209), anti-CD24-BrilliantViolet 510 (Biolegend, cat# 101831) and anti-IgD-PE (Biolegend, cat# 405705). For the subpopulations quantification, cells within a B220+/IgM- gate, were characterized for expression of CD43 and CD24, and defined as: Pre -Pro-B (CD43+, CD24-/low), Pro-B (CD43+, CD24+), Pre-B (CD43-/CD24+). A second gate of B220+ cells was used to quantify immature B cells (IgM+, IgD) or transitional B cells (IgM high, IgD low). In all analyses, blocking was performed with rat anti-mouse CD16/CD32 (Biolegend, cat# 101302), propidium iodide staining was used for live cells gating, and compensation performed with UltraComp eBeads™ (Invitrogen, cat# 01–2222-42). Cells were acquired using a BD LSRII equipped with BD FACSDiva v8.0.1. (BD Bioscience, San Jose, CA, USA). Data was analyzed with FlowJo software v10.6.2 (FlowJo LLC, Ashlang, OR, USA)

Southern blot.

High molecular weight DNA was obtained from frozen spleens of Eμ-Irf8 and WT mice (Gentra Puregene, Qiagen, cat# 158467). For detection of the IGH germline locus, DNA was isolated from murine kidney. Ten micrograms of DNA were digested with EcoRI for 24h at 37°C. Digested DNA was separated in 0.8% agarose gel and transferred to a nylon membrane, according to standard procedures[30].The membrane was hybridized with the PerfectHyb Plus buffer (Sigma-Aldrich, cat# H7033) for 16h at 65°C using a ³²P-labeled PCR-generated 0.65Kb probe spanning the murine 3’ Jh region[31].

Histopathology studies

Murine spleens and cervical lymph nodes were fixed in formalin, embedded in paraffin, and 4-μm serial sections obtained, as we described[29]. One section was stained with Hematoxylin and Eosin (H&E) and the slides analyzed by two pathologists (KNH and MCK), who were blind to the mice genotype. Automated immunohistochemical stains were performed on Ventana Discovery (Ventana Medical Systems), following antigen retrieval in CC1, with rat monoclonal anti-B220 (BD Biosciences clone RA3–6B2, 1:200) and rabbit monoclonal anti-CD3 (ThermoFisher, clone SP7, 1:200). All immunohistochemical sections were counterstained with hematoxylin.

Statistical analyses

P values were calculated with one-way ANOVA with Tukey’s multiple comparisons test, Student’s t-test (two-tailed, equal variance), or Chi-square test. Survival proportions were calculated using the Kaplan-Meier method. Statistical significance of survival estimates was calculated using the log-rank (Mantel-Cox) test; Hazard ratio and confidence interval were also determined. All statistical analyses wer e completed using the GraphPad Prism 8 software (GraphPad Software). P value <0.05 was considered significant.

Results

Generation of the Eμ-Irf8 mouse model

Full length murine Irf8 driven by the Eμ enhancer and V186.2 promoter (Figure 1a) was injected into one-cell C57BL/6 embryos. From these pronuclear injections, two litters totaling 15 pups were born, and three founder animals identified; each founder was independently backcrossed into C57BL/6 mice and three lines of mice, referred to as Irf8–100, Irf8–200 and Irf8–300 mice established. Eμ-Irf8 mice from all three lines were viable, fertile, and displayed normal development, irrespective of whether the mice were the hemizygous or homozygous to the Eμ-Irf8 allele. Transgene quantification in the founder mice indicated that 3 to 9 Irf8 copies integrated in each genome (Figure 1b). Next, we isolated mature B cells from 4-month old Eμ-Irf8 or WT mice and used a transgene-specific RT-PCR to confirm ectopic Irf8 expression in these cells (Figure 1c); in addition, using q-RT-PCR in we confirmed that Irf8’s expression was approximately twice higher in Eμ-Irf8 mice than WT littermate controls (Figure 1c). Note that the first RT-PCR in this panel, only detects the transgene, hence the lack of expression in WT mice, whereas the q-RT-PCR detects both endogenous and ectopic Irf8. Higher expresison of Irf8 in the Eμ-Irf8 mice was validated at protein level using western blot (WB) in young (6 months) as well older mice (12 to 18-month old) (Figure 1d). Importantly, demostrating the functional relevance of Irf8’s aberrant expression, the levels of its direct targets Aicda and Bcl6, were elevated Eμ-Irf8 in comparison to WT control (Figure 1d). We concluded that the Eμ-Irf8 transgene allele did not impact the normal development of mice in three distinct founder lines, that the degree of overexpression of Irf8 in these mice was not overtly elevated and closely reflects the overexpression noted in the primary DLBCL with the IGH-IRF8 fusion[5], and that it increased the expression of two IRF8 targets that are relevant to B cell lymphomagenesis.

Figure 1. Generation and initial characterization of the Eμ-Irf8 mouse model.

A. Schematic diagram of the Irf8 transgenic construct. B. PCR-based determination of the Irf8 transgene copy number in three founders lines. Densitometric estimation of the number of Irf8 copies is 3, 8 and 9 for lines 100, 200 and 300, respectively. C. Left - A transgene-specific RT-PCR confirmed the ectopic expression of Irf8 in the mature B-cells from the three founder lines. The lack of amplification in the “no RT” samples, confirmed that the VH-Irf8product does not derive from DNA contaminating the RNA prep. Right - Irf8 q-RT-PCR in spleen from 4 months old Eμ-Irf8 and WT littermates; data are mean -+SD, statistical test in one-way ANOVA. D. Western blot analysis of Irf8, Aicda and Bcl6 in spleen from young (left) or old (right) WT and Eμ-Irf8 littermates. Irf8 and Aicda were analyzed in the same WB and share the β-actin.

Hematological features of the Eμ-Irf8 mice

To examine the potential impact of the Eμ-Irf8 transgene in the hematological lineage, we performed serial complete blood counts (CBC) in eight mice of the original litter, including the 3 Eμ-Irf8 founders and 5 WT littermates. We did not detect any significant difference in the counts between these two cohorts at 3 or 6 months of age (Figure 2a). To expand on these data, we performed CBC in a larger cohort of older mice (mean = 15 months, range, 11 to 18) that included WT, Eμ-Irf8 hemizygous or homozygous mice (n=17). Again, we did not detect any significant difference across these genotypes (Figure 2b). We also examined total bone marrow and spleen cellularity and found no significant differences among sex and age-matched WT or Eμ-Irf8 mice (n=20) (Figure 2c). In agreement with these observations, the size and weight of spleens from large cohorts (n=55 and n=41, respectively) of WT and Eμ-Irf8 mice, matched for age and sex, were not significantly different (Figure 2c). We concluded the expression of the Eμ-Irf8 transgene does not modify broad hematological parameters in mice.

Figure 2. Hematological examination of the Eμ-Irf8 mouse model.

A. Complete blood count in the Eμ-Irf8 founder’s litter at 3 months (n=8, 5 WT and 3 Eμ-Irf8 mice) and at 6 months of age (n=5, 3 WT and 2 Eμ-Irf8 mice). B. Complete blood count of older (mean = 15 months old, range 11 to 18 months) WT and Eμ-Irf8 mice (hemizygous and homozygous) (n=17). C. Left panels - Bone marrow and spleen cellularity in WT and Eμ-Irf8 spanning multiple age groups (n=10, 8–12 weeks old; n=12, 15–18 months old); Right panels - spleen weight and size in WT and Eμ-Irf8 mice. All data shown are mean +-SEM; statistical tests were one-way ANOVA with Tukey’s multiple comparisons test (panels A and B), and two=tailed Student’s t-test (panel C).

Characterization of B cell subpopulations in the spleen and bone marrow of Eμ-Irf8 mice

To detect the influence of Irf8 in discrete B cell sub-compartments, we performed FACS analysis in primary and secondary lymphoid organs, bone marrow (BM) and spleen, in cohorts of young (median 9-week old, n=10) and older (median 17-month old, n=10) WT and Eμ-Irf8 mice representative of all founder lines. In the bone marrow, we charac terized five B cell subpopulations - pre-pro-B, pro-B, pre-B, immature and transitional B cells. In young mice,the number of total B cells (B220+) as well as the representation of their subsets was similar in WT and Eμ-Irf8 mice (Figure 3a). Conversely, in the older cohort, there was a trend for fewer B cells in BM of Eμ-Irf8 than in WT mice (mean 20% vs. 33%, p=0.06, ns, Figure 3a), accompanied by accumulation of pre-pro-B and fewer pre-B cells, although these differences did not reach statistical significance. Importantly, we confirmed by WB higher expression of Irf8 in the bone marrow of Eμ-Irf8 mice in comparison to WT controls (Figure 3a). Next, we characterized the B cell subpopulations in the spleens of young and older Eμ-Irf8 mice. The percentage all B cells (B220+) in the spleen, as well as of transitional, follicular and marginal zone sub-sets were similar in young mice, irrespective of the presence of the Irf8 transgene (Figure 3b). Likewise, in the older cohort no statistically significant differences were detected, although a trend for accumulation of transitional B cells was detected in Eμ-Irf8 mice (mean 12.4% vs. 4.7%,p=ns) (Figure 3b). We concluded that expression of the Eμ -Irf8 transgene does not significantly modify the representation of B cell sub-compartments in primary or secondary lymphoid organs of young mice, but it may induce modest changes in older mice, notably a decrease in the B cell pool in the BM and accumulation of transitional B cells in the spleen.

Figure 3. Characterization of primary and secondary lymphoid organs in the Eμ-Irf8 mouse model.

A. Left to right panels: FACS analysis of B-cell sub-populations in the bone marrow of young (8–12 weeks) or older (15–18 months) WT and Eμ-Irf8 mice (n=20). Representative FACS plots depicting the trend for a lower percentage of B220+ cells in the BM of older Eμ-Irf8 mice is shown to right. WB shows the higher expression of Irf8 in the BM of Eμ-Irf8 mice. B. Left to right panels: FACS analysis of B-cell sub-populations in the spleen of young (8–12 weeks) or older (15–18 months) WT and Eμ-Irf8 mice (n=20). Representative FACS plots depicting the trend for a higher percentage of transitional B cells in the spleen of Eμ-Irf8 mice is shown to right. Follicular B cells are colored in burgundy, marginal zone B cells in green and transitional B cells in black.

Survival of aged Eμ-Irf8 mice

Eμ-Irf8 mice are viable, fertile, and display normal development. Furthemore, these mice do not develop clinically detectable lymphoma or display overtly distinct survival rate in comparison to WT controls in their first 12–18 months of life. For these reasons, we aged large cohorts of WT (n=78) and Eμ-Irf8 mice (n=78 - representative of the three founder lines) and estimated their survival probability. We found that the median survival of Eμ-Irf8 mice was significantly shorter than that of WT controls (23.2 vs. 28.5 months, p=0.02, log-rank Mantel Cox test, hazard ratio [logrank] = 2.3, 95% CI 1.1 – 4.9) (Figure 4). We concluded that deregulated expression of Irf8 in B cells (via the Eμ-Irf8 transgene) shortened mice survival, a difference that becomes more apparent beyond 12 to 18 months of life.

Figure 4. Survival analysis of the Eμ-Irf8 mouse.

Survival estimates of WT and Eμ-Irf8 mice (n=156) were calculated using the Kaplan-Meier method. Mice with overexpression of Irf8 had a significantly shorter median survival (23.2 vs. 28.5 months). Statistical significance of survival proportions was calculated using the log-rank (Mantel-Cox) test.

Histopathological and molecular characterization of secondary Iymphoid organs in WT and Eμ-Irf8 mice

Analysis of B cell subpopulations in the spleen and BM and survival probability curves suggested that similarly to other mouse models of B cell lymphoma (e.g., Bcl6, Tet1, kmt2d)[32–34], a phenotype derived from Irf8 overexpression may emerge only in older mice. To address this possibility, we collected spleen and lymph-nodes from 38 Eμ-Irf8 and WT mice with a median age of 25 months (range 20 to 35) and equal sex distribution (19 male, 19 female). Importantly, considering the known incidence of lympho-proliferations in aged C57BL/6J mice[35], and the need to include a representative number of Eμ-Irf8 mice from the three founder lines, this cohort purposedly included a ~ 3 fold excess of Eμ-Irf8 mice (n=29 vs. 9 WT). Histopathological examination of the spleens (n=38) or lymph -nodes (n=30) was completed independently by two hemato-pathologists (KNH and MCK), and molecular analysis of B cell clonality was performed by Southern blot in all spleens for which sufficient DNA was available (n=32). Ten of the 29 Eμ-Irf8 tissues analyzed (34%) and three of the 9 WT spleens/LN (33%) displayed features of “lymphoproliferative disorder” (LPD), herein broadly used to describe features compatible lymphoid hyperplasia or early stages of follicular lymphoma (follicular hyperplasia), or overt follicular lymphoma (FL), marginal zone lymphoma (MZL), and DLBCL (Figure 5a). Southern blot analysis of the immunoglobulin heavy chain locus documented a clonal LDP/FL/DLBCL/MZL in 3 of 8 WT and 8 of 24 Eμ-Irf8 tumors (p=ns, Chi-square) (Figure 5b). The inability to detect clonal rearrangements in a small fraction of the abnormal tissues examined is likely secondary to the limited representation of tumor cells in cases labeledas lymphoid hyperplasia/follicular hyperplasia.

Figure 5. Histopathological and molecular analysis of the aged WT and Eμ-Irf8 mice.

A. Top. Prevalence of B cell lymphoproliferative diseases in mice of the indicated genotypes. Color-coded segments correspond to lymphoid tissue within the normal limits, tissue with features of lymphoproliferation/hyperplasia but not overt lymphoma, or with classical follicular or diffuse large B cell lymphoma. The number of animals in the normal category is given inside the bar, and the overall number of animals diagnosed with LPD/lymphoma out of the total analyzed is above the bars. Bottom. Histologic and immunohistochemical analysis of representative spleens from control WT mouse, left most panels, and three Eμ-Irf8 mice diagnosed with FL low-grade, FL high-grade/DLBCL and MZL. Top, H&E staining 4X, middle, H&E staining 40X, Bottom, imunostaining for B220 in cases with abnormal histology. B. Top. Color-coded segments indicated the number of clonal lymphoproliferations in mice of the indicated genotypes. Bottom. Southern blot analysis of EcoRI-digested DNA from representative spleens. Kidney tissue served as a control for the germline immunoglobulin heavy chain (GL control ~6.5Kb), spleen of an Eμ-Myc mouse served as positive control for a clonal rearrangement (red arrow); one spleen (lymphoma) each from a Eμ-Irf8or from a WT mouse show a clonally rearranged IGH locus (red arrow). Statistical analysis was performed with the Chi-square test.

Discussion

The transcription factor IRF8 displays several features that align with other cancer-associated genes. For example, according to the target tissue, WT IRF8 expression can suppress[11, 12] or promote oncogenesis[5, 7, 36], and it can be deregulated by multiple modes, including elevated expression, somatic mutations or chromosomal translocations[5–7, 22]. In brief, in the myeloid lineage, IRF8 functions as “tumor-suppressor gene” as initially demonstrated by CML-like disease in Irf8 KO mice[11]. In addition, BCR-ABL mediates IRF8 suppression in CML, increasing the potential of leukemia-initiating cells and promoting imatinib resistance[37–39]. Conversely, in the B lymphoid lineage, IRF8 has the hallmarks of an oncogene. It is targeted by missense variants clustered in key functional domains in ~10% of DLBCL[6, 7]. More importantly, we showed that through a balanced chromosomal translocation the IRF8 locus is juxtaposed to the IGH regulatory elements[5], a genetic deregulation that typifies mature B cell cancers[2]. While the study of the interplay between IRF8 and myeloid malignancies has benefited from the availability of the Irf8 KO mice, a model relevant to the examination of effects of Irf8 gain in lymphoid malignancies has been lacking. Here, we described the generation and characterization of the Eμ-Irf8 mouse, which models the t(14;16) (q32;q24) and genomic IGH-IRF8 fusion found in subsets of DLBCL[5, 22]. Examining three independent founder lines, we showed that Eμ-Irf8 mice are viable, born at normal Mendelian and gender ratios, develop normally, but have a shorter lifespan than their WT littermates (23 vs. 28 months).

IRF8 promotes the development of pre-pro-B cells in the BM[40]. In addition, IRF8, together with IRF4, PU.1 and SpiB are essential for Ig light-chain gene expression and the generation of immature B cells in the BM[40, 41], which migrate to the spleen, where they continue to differentiate into transitional B cells. In agreement with these biological precepts, when examining B cell subpopulations in primary and secondary lymphoid organs, we found accumulation of pre-pro-B cells in the BM and transitional B cells in the spleen of Eμ-Irf8 mice. In addition, mature B cells from the Eμ-Irf8 mouse consistently displayed higher expression of Aicda and Bcl6, two well-characterized IRF8 targets[17] that have well-defined lymphomagenic activities[32, 42, 43]. These data give confidence that the Eμ-Irf8 recapitulates Irf8 biology in vivo. Yet, intriguingly, the Eμ-Irf8 mice did not display an overt phenotype. In this respect, it is important to consider that single knockout models of either Irf8 or its common DNA binding PU.1, in CD19+ B cells also resulted in minimal phenotype, probably reflecting a certain degree of redundancy with closely related members of their respective transcription factor families[14, 44, 45].

Interestingly, the Eμ-Irf8 mice displayed a shorter lifespan than WT controls, which could not be attributed to the development of B cell malignancies. The reason(s) for this discrepancy are at the moment unclear. It is possible that low penetrance and late tumor development, common features of mouse models of lymphoma (e.g., B cell conditional Bcl6 knock-in[32] and Tet1 knockout[34]), may play a role. Indeed, considering the large cohort size (n =156) utilized for determining the shorter of survival probability of the Eμ-Irf8 mice, a power calculation estimation indicates that to detect a small (~20%) but significant difference in B cell lymphoma incidence between Eμ-Irf8 and WT mice, will require analysis of close to 175 aged(>24-month old) mice. Unfortunately, such large cohort is still not available, but we continue to accrue mice in the “histopathology cohort” as they pass the 2-year old mark, and thus should be able to in the future fully test the low penetrance hypothesis. Notably, the trend for an abnormal B cell development in the spleen of Eμ-Irf8 (accumulation of transitional B cells) may be a harbinger for a more established defect to emerge upon histological examination of very large cohorts.

It also important to consider that the Irf8 expression in the Eμ-Irf8 mice is only moderately higher than that of WT mice, a feature that agrees to the findings in DLBCL with t(14;16)[5]. This notion is relevant because in a recent report on the role of Irf8 in myeloid-derived suppressor cells, to rescue a KO phenotype, transgenes with a six-fold higher Irf8 expression were needed[46]. However, we argue that while our strategy may have decreased the penetrance or delayed lymphoma development, it more faithfully recapitulates IRF8 deregulation in human DLBCL and thus eventually it may yield more informative data. In spite of these caveats, presently we cannot exclude the possibility that the shorter survival of the Eμ-Irf8 is unrelated to a putative hematological/B cell dysfunction, although gross anatomy of the mice subject histopathology did not uncover any obvious abnormality.

In humans, IRF8 translocation and mutations are more common in the germinal-center like (GCB)-like subtype of DLBCL[15, 22], but also detected in a significant fraction in activated B-cell (ABC)-like and unclassifiable DBCBLs[7]. In agreement with its broader role, a CRISPR-screening showed that IRF8 behaved as a “variable” essential gene, that is, its deletion could inducecell death in GCB- and ABC-DLBCL cell lines alike[7]. Importantly, recent and more extensive examination of the molecular heterogeneity of DLBCL, mapped IRF8 mutants tumors to a cluster of tumors (C3) that also display BCL2 translocation, and mutation in the chromatin modifiers KMT2D, CREBBP and EZH2[6]. Although lymphomas in this cluster belonged predominantly to the GCB-DLBCL subtype, these patients displayed poor outcome, providing further evidence of the limitation of the GCB vs. ABC dichotomy as a reliable outcome predictor in DLBCL[3]. In the context of the Eμ -Irf8 mouse model, these data highlight candidate genes that when disrupted in a background of Irf8 overexpression could fully uncover the contribution of this transcription factor to B-cell lymphoma biology. Indeed, full penetrance and early disease development in most mouse models of lymphoma often require at least two genetic “hits”, with the Eμ-Myc model, which spontaneously acquires second hits in the p53/p16/p19Arf pathway, as a notable exception[47]. Thus, immediately available strategies are to cross Eμ-Irf8 mice with VavP-Bcl2[48], Ezh2 conditional knock-in[49],or Kmt2d[33] and Crebbp[50] conditional knockout mice, all carrying genetic lesions that cluster with IRF8 deregulation in human DLBCL[6]. In fact, a similar approach was recently successfully employed to unveil the lymphomagenic role of Kmt2d and Crebbp in B cell lymphoma, by crossing mice with conditional knockout of these chromatin modifiers with the VavP-Bcl2 model[33, 50]. Thus, we postulate that analysis of larger cohorts of aged mice and/or introduction of biologically relevant second “hits” willbe needed to uncover the full phenotype of Eμ-Irf8 mouse model.

In summary, herein we described the generation and characterization of the first mouse model of Irf8 overexpression in B cells driven by the immunoglobulin regulatory sequences. We found that the Eμ-Irf8 mice have a shorter lifespan than WT littermates, and likely develop lymphoproliferation at low penetrance, which emerge at advanced age (>24–30 months). We suggest that deregulated Irf8 expression may be insufficient to initiate malignant transformation, and that compound Eμ-Irf8/Bcl2-transgenic mice, Eμ-Irf8/Kmt2d-ko or Eμ-Irf8/Crebbp-ko mice, which faithfully recapitulate the genetics of human DLBCL, should be generated in the future.

Data availability:

The data that support the findings of this study are available in the main manuscript, supplementary materials and can be requested from the corresponding author.